2 * linux/fs/ext4/xattr.c
4 * Copyright (C) 2001-2003 Andreas Gruenbacher, <agruen@suse.de>
6 * Fix by Harrison Xing <harrison@mountainviewdata.com>.
7 * Ext4 code with a lot of help from Eric Jarman <ejarman@acm.org>.
8 * Extended attributes for symlinks and special files added per
9 * suggestion of Luka Renko <luka.renko@hermes.si>.
10 * xattr consolidation Copyright (c) 2004 James Morris <jmorris@redhat.com>,
12 * ea-in-inode support by Alex Tomas <alex@clusterfs.com> aka bzzz
13 * and Andreas Gruenbacher <agruen@suse.de>.
17 * Extended attributes are stored directly in inodes (on file systems with
18 * inodes bigger than 128 bytes) and on additional disk blocks. The i_file_acl
19 * field contains the block number if an inode uses an additional block. All
20 * attributes must fit in the inode and one additional block. Blocks that
21 * contain the identical set of attributes may be shared among several inodes.
22 * Identical blocks are detected by keeping a cache of blocks that have
23 * recently been accessed.
25 * The attributes in inodes and on blocks have a different header; the entries
26 * are stored in the same format:
28 * +------------------+
31 * | entry 2 | | growing downwards
36 * | value 3 | | growing upwards
38 * +------------------+
40 * The header is followed by multiple entry descriptors. In disk blocks, the
41 * entry descriptors are kept sorted. In inodes, they are unsorted. The
42 * attribute values are aligned to the end of the block in no specific order.
46 * EXT4_I(inode)->i_file_acl is protected by EXT4_I(inode)->xattr_sem.
47 * EA blocks are only changed if they are exclusive to an inode, so
48 * holding xattr_sem also means that nothing but the EA block's reference
49 * count can change. Multiple writers to the same block are synchronized
53 #include <linux/init.h>
55 #include <linux/slab.h>
56 #include <linux/mbcache.h>
57 #include <linux/quotaops.h>
58 #include <linux/rwsem.h>
59 #include "ext4_jbd2.h"
64 #define BHDR(bh) ((struct ext4_xattr_header *)((bh)->b_data))
65 #define ENTRY(ptr) ((struct ext4_xattr_entry *)(ptr))
66 #define BFIRST(bh) ENTRY(BHDR(bh)+1)
67 #define IS_LAST_ENTRY(entry) (*(__u32 *)(entry) == 0)
69 #ifdef EXT4_XATTR_DEBUG
70 # define ea_idebug(inode, f...) do { \
71 printk(KERN_DEBUG "inode %s:%lu: ", \
72 inode->i_sb->s_id, inode->i_ino); \
76 # define ea_bdebug(bh, f...) do { \
77 char b[BDEVNAME_SIZE]; \
78 printk(KERN_DEBUG "block %s:%lu: ", \
79 bdevname(bh->b_bdev, b), \
80 (unsigned long) bh->b_blocknr); \
85 # define ea_idebug(f...)
86 # define ea_bdebug(f...)
89 static void ext4_xattr_cache_insert(struct buffer_head *);
90 static struct buffer_head *ext4_xattr_cache_find(struct inode *,
91 struct ext4_xattr_header *,
92 struct mb_cache_entry **);
93 static void ext4_xattr_rehash(struct ext4_xattr_header *,
94 struct ext4_xattr_entry *);
95 static int ext4_xattr_list(struct inode *inode, char *buffer,
98 static struct mb_cache *ext4_xattr_cache;
100 static struct xattr_handler *ext4_xattr_handler_map[] = {
101 [EXT4_XATTR_INDEX_USER] = &ext4_xattr_user_handler,
102 #ifdef CONFIG_EXT4_FS_POSIX_ACL
103 [EXT4_XATTR_INDEX_POSIX_ACL_ACCESS] = &ext4_xattr_acl_access_handler,
104 [EXT4_XATTR_INDEX_POSIX_ACL_DEFAULT] = &ext4_xattr_acl_default_handler,
106 [EXT4_XATTR_INDEX_TRUSTED] = &ext4_xattr_trusted_handler,
107 #ifdef CONFIG_EXT4_FS_SECURITY
108 [EXT4_XATTR_INDEX_SECURITY] = &ext4_xattr_security_handler,
112 struct xattr_handler *ext4_xattr_handlers[] = {
113 &ext4_xattr_user_handler,
114 &ext4_xattr_trusted_handler,
115 #ifdef CONFIG_EXT4_FS_POSIX_ACL
116 &ext4_xattr_acl_access_handler,
117 &ext4_xattr_acl_default_handler,
119 #ifdef CONFIG_EXT4_FS_SECURITY
120 &ext4_xattr_security_handler,
125 static inline struct xattr_handler *
126 ext4_xattr_handler(int name_index)
128 struct xattr_handler *handler = NULL;
130 if (name_index > 0 && name_index < ARRAY_SIZE(ext4_xattr_handler_map))
131 handler = ext4_xattr_handler_map[name_index];
136 * Inode operation listxattr()
138 * dentry->d_inode->i_mutex: don't care
141 ext4_listxattr(struct dentry *dentry, char *buffer, size_t size)
143 return ext4_xattr_list(dentry->d_inode, buffer, size);
147 ext4_xattr_check_names(struct ext4_xattr_entry *entry, void *end)
149 while (!IS_LAST_ENTRY(entry)) {
150 struct ext4_xattr_entry *next = EXT4_XATTR_NEXT(entry);
151 if ((void *)next >= end)
159 ext4_xattr_check_block(struct buffer_head *bh)
163 if (BHDR(bh)->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC) ||
164 BHDR(bh)->h_blocks != cpu_to_le32(1))
166 error = ext4_xattr_check_names(BFIRST(bh), bh->b_data + bh->b_size);
171 ext4_xattr_check_entry(struct ext4_xattr_entry *entry, size_t size)
173 size_t value_size = le32_to_cpu(entry->e_value_size);
175 if (entry->e_value_block != 0 || value_size > size ||
176 le16_to_cpu(entry->e_value_offs) + value_size > size)
182 ext4_xattr_find_entry(struct ext4_xattr_entry **pentry, int name_index,
183 const char *name, size_t size, int sorted)
185 struct ext4_xattr_entry *entry;
191 name_len = strlen(name);
193 for (; !IS_LAST_ENTRY(entry); entry = EXT4_XATTR_NEXT(entry)) {
194 cmp = name_index - entry->e_name_index;
196 cmp = name_len - entry->e_name_len;
198 cmp = memcmp(name, entry->e_name, name_len);
199 if (cmp <= 0 && (sorted || cmp == 0))
203 if (!cmp && ext4_xattr_check_entry(entry, size))
205 return cmp ? -ENODATA : 0;
209 ext4_xattr_block_get(struct inode *inode, int name_index, const char *name,
210 void *buffer, size_t buffer_size)
212 struct buffer_head *bh = NULL;
213 struct ext4_xattr_entry *entry;
217 ea_idebug(inode, "name=%d.%s, buffer=%p, buffer_size=%ld",
218 name_index, name, buffer, (long)buffer_size);
221 if (!EXT4_I(inode)->i_file_acl)
223 ea_idebug(inode, "reading block %u", EXT4_I(inode)->i_file_acl);
224 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
227 ea_bdebug(bh, "b_count=%d, refcount=%d",
228 atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
229 if (ext4_xattr_check_block(bh)) {
230 bad_block: ext4_error(inode->i_sb, __func__,
231 "inode %lu: bad block %llu", inode->i_ino,
232 EXT4_I(inode)->i_file_acl);
236 ext4_xattr_cache_insert(bh);
238 error = ext4_xattr_find_entry(&entry, name_index, name, bh->b_size, 1);
243 size = le32_to_cpu(entry->e_value_size);
246 if (size > buffer_size)
248 memcpy(buffer, bh->b_data + le16_to_cpu(entry->e_value_offs),
259 ext4_xattr_ibody_get(struct inode *inode, int name_index, const char *name,
260 void *buffer, size_t buffer_size)
262 struct ext4_xattr_ibody_header *header;
263 struct ext4_xattr_entry *entry;
264 struct ext4_inode *raw_inode;
265 struct ext4_iloc iloc;
270 if (!(EXT4_I(inode)->i_state & EXT4_STATE_XATTR))
272 error = ext4_get_inode_loc(inode, &iloc);
275 raw_inode = ext4_raw_inode(&iloc);
276 header = IHDR(inode, raw_inode);
277 entry = IFIRST(header);
278 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
279 error = ext4_xattr_check_names(entry, end);
282 error = ext4_xattr_find_entry(&entry, name_index, name,
283 end - (void *)entry, 0);
286 size = le32_to_cpu(entry->e_value_size);
289 if (size > buffer_size)
291 memcpy(buffer, (void *)IFIRST(header) +
292 le16_to_cpu(entry->e_value_offs), size);
304 * Copy an extended attribute into the buffer
305 * provided, or compute the buffer size required.
306 * Buffer is NULL to compute the size of the buffer required.
308 * Returns a negative error number on failure, or the number of bytes
309 * used / required on success.
312 ext4_xattr_get(struct inode *inode, int name_index, const char *name,
313 void *buffer, size_t buffer_size)
317 down_read(&EXT4_I(inode)->xattr_sem);
318 error = ext4_xattr_ibody_get(inode, name_index, name, buffer,
320 if (error == -ENODATA)
321 error = ext4_xattr_block_get(inode, name_index, name, buffer,
323 up_read(&EXT4_I(inode)->xattr_sem);
328 ext4_xattr_list_entries(struct inode *inode, struct ext4_xattr_entry *entry,
329 char *buffer, size_t buffer_size)
331 size_t rest = buffer_size;
333 for (; !IS_LAST_ENTRY(entry); entry = EXT4_XATTR_NEXT(entry)) {
334 struct xattr_handler *handler =
335 ext4_xattr_handler(entry->e_name_index);
338 size_t size = handler->list(inode, buffer, rest,
349 return buffer_size - rest;
353 ext4_xattr_block_list(struct inode *inode, char *buffer, size_t buffer_size)
355 struct buffer_head *bh = NULL;
358 ea_idebug(inode, "buffer=%p, buffer_size=%ld",
359 buffer, (long)buffer_size);
362 if (!EXT4_I(inode)->i_file_acl)
364 ea_idebug(inode, "reading block %u", EXT4_I(inode)->i_file_acl);
365 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
369 ea_bdebug(bh, "b_count=%d, refcount=%d",
370 atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
371 if (ext4_xattr_check_block(bh)) {
372 ext4_error(inode->i_sb, __func__,
373 "inode %lu: bad block %llu", inode->i_ino,
374 EXT4_I(inode)->i_file_acl);
378 ext4_xattr_cache_insert(bh);
379 error = ext4_xattr_list_entries(inode, BFIRST(bh), buffer, buffer_size);
388 ext4_xattr_ibody_list(struct inode *inode, char *buffer, size_t buffer_size)
390 struct ext4_xattr_ibody_header *header;
391 struct ext4_inode *raw_inode;
392 struct ext4_iloc iloc;
396 if (!(EXT4_I(inode)->i_state & EXT4_STATE_XATTR))
398 error = ext4_get_inode_loc(inode, &iloc);
401 raw_inode = ext4_raw_inode(&iloc);
402 header = IHDR(inode, raw_inode);
403 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
404 error = ext4_xattr_check_names(IFIRST(header), end);
407 error = ext4_xattr_list_entries(inode, IFIRST(header),
408 buffer, buffer_size);
418 * Copy a list of attribute names into the buffer
419 * provided, or compute the buffer size required.
420 * Buffer is NULL to compute the size of the buffer required.
422 * Returns a negative error number on failure, or the number of bytes
423 * used / required on success.
426 ext4_xattr_list(struct inode *inode, char *buffer, size_t buffer_size)
428 int i_error, b_error;
430 down_read(&EXT4_I(inode)->xattr_sem);
431 i_error = ext4_xattr_ibody_list(inode, buffer, buffer_size);
437 buffer_size -= i_error;
439 b_error = ext4_xattr_block_list(inode, buffer, buffer_size);
443 up_read(&EXT4_I(inode)->xattr_sem);
444 return i_error + b_error;
448 * If the EXT4_FEATURE_COMPAT_EXT_ATTR feature of this file system is
451 static void ext4_xattr_update_super_block(handle_t *handle,
452 struct super_block *sb)
454 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_EXT_ATTR))
457 if (ext4_journal_get_write_access(handle, EXT4_SB(sb)->s_sbh) == 0) {
458 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_EXT_ATTR);
460 ext4_journal_dirty_metadata(handle, EXT4_SB(sb)->s_sbh);
465 * Release the xattr block BH: If the reference count is > 1, decrement
466 * it; otherwise free the block.
469 ext4_xattr_release_block(handle_t *handle, struct inode *inode,
470 struct buffer_head *bh)
472 struct mb_cache_entry *ce = NULL;
475 ce = mb_cache_entry_get(ext4_xattr_cache, bh->b_bdev, bh->b_blocknr);
476 error = ext4_journal_get_write_access(handle, bh);
481 if (BHDR(bh)->h_refcount == cpu_to_le32(1)) {
482 ea_bdebug(bh, "refcount now=0; freeing");
484 mb_cache_entry_free(ce);
485 ext4_free_blocks(handle, inode, bh->b_blocknr, 1, 1);
487 ext4_forget(handle, 1, inode, bh, bh->b_blocknr);
489 le32_add_cpu(&BHDR(bh)->h_refcount, -1);
490 error = ext4_journal_dirty_metadata(handle, bh);
493 DQUOT_FREE_BLOCK(inode, 1);
494 ea_bdebug(bh, "refcount now=%d; releasing",
495 le32_to_cpu(BHDR(bh)->h_refcount));
497 mb_cache_entry_release(ce);
501 ext4_std_error(inode->i_sb, error);
506 * Find the available free space for EAs. This also returns the total number of
507 * bytes used by EA entries.
509 static size_t ext4_xattr_free_space(struct ext4_xattr_entry *last,
510 size_t *min_offs, void *base, int *total)
512 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
513 *total += EXT4_XATTR_LEN(last->e_name_len);
514 if (!last->e_value_block && last->e_value_size) {
515 size_t offs = le16_to_cpu(last->e_value_offs);
516 if (offs < *min_offs)
520 return (*min_offs - ((void *)last - base) - sizeof(__u32));
523 struct ext4_xattr_info {
530 struct ext4_xattr_search {
531 struct ext4_xattr_entry *first;
534 struct ext4_xattr_entry *here;
539 ext4_xattr_set_entry(struct ext4_xattr_info *i, struct ext4_xattr_search *s)
541 struct ext4_xattr_entry *last;
542 size_t free, min_offs = s->end - s->base, name_len = strlen(i->name);
544 /* Compute min_offs and last. */
546 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
547 if (!last->e_value_block && last->e_value_size) {
548 size_t offs = le16_to_cpu(last->e_value_offs);
553 free = min_offs - ((void *)last - s->base) - sizeof(__u32);
555 if (!s->here->e_value_block && s->here->e_value_size) {
556 size_t size = le32_to_cpu(s->here->e_value_size);
557 free += EXT4_XATTR_SIZE(size);
559 free += EXT4_XATTR_LEN(name_len);
562 if (free < EXT4_XATTR_SIZE(i->value_len) ||
563 free < EXT4_XATTR_LEN(name_len) +
564 EXT4_XATTR_SIZE(i->value_len))
568 if (i->value && s->not_found) {
569 /* Insert the new name. */
570 size_t size = EXT4_XATTR_LEN(name_len);
571 size_t rest = (void *)last - (void *)s->here + sizeof(__u32);
572 memmove((void *)s->here + size, s->here, rest);
573 memset(s->here, 0, size);
574 s->here->e_name_index = i->name_index;
575 s->here->e_name_len = name_len;
576 memcpy(s->here->e_name, i->name, name_len);
578 if (!s->here->e_value_block && s->here->e_value_size) {
579 void *first_val = s->base + min_offs;
580 size_t offs = le16_to_cpu(s->here->e_value_offs);
581 void *val = s->base + offs;
582 size_t size = EXT4_XATTR_SIZE(
583 le32_to_cpu(s->here->e_value_size));
585 if (i->value && size == EXT4_XATTR_SIZE(i->value_len)) {
586 /* The old and the new value have the same
587 size. Just replace. */
588 s->here->e_value_size =
589 cpu_to_le32(i->value_len);
590 memset(val + size - EXT4_XATTR_PAD, 0,
591 EXT4_XATTR_PAD); /* Clear pad bytes. */
592 memcpy(val, i->value, i->value_len);
596 /* Remove the old value. */
597 memmove(first_val + size, first_val, val - first_val);
598 memset(first_val, 0, size);
599 s->here->e_value_size = 0;
600 s->here->e_value_offs = 0;
603 /* Adjust all value offsets. */
605 while (!IS_LAST_ENTRY(last)) {
606 size_t o = le16_to_cpu(last->e_value_offs);
607 if (!last->e_value_block &&
608 last->e_value_size && o < offs)
610 cpu_to_le16(o + size);
611 last = EXT4_XATTR_NEXT(last);
615 /* Remove the old name. */
616 size_t size = EXT4_XATTR_LEN(name_len);
617 last = ENTRY((void *)last - size);
618 memmove(s->here, (void *)s->here + size,
619 (void *)last - (void *)s->here + sizeof(__u32));
620 memset(last, 0, size);
625 /* Insert the new value. */
626 s->here->e_value_size = cpu_to_le32(i->value_len);
628 size_t size = EXT4_XATTR_SIZE(i->value_len);
629 void *val = s->base + min_offs - size;
630 s->here->e_value_offs = cpu_to_le16(min_offs - size);
631 memset(val + size - EXT4_XATTR_PAD, 0,
632 EXT4_XATTR_PAD); /* Clear the pad bytes. */
633 memcpy(val, i->value, i->value_len);
639 struct ext4_xattr_block_find {
640 struct ext4_xattr_search s;
641 struct buffer_head *bh;
645 ext4_xattr_block_find(struct inode *inode, struct ext4_xattr_info *i,
646 struct ext4_xattr_block_find *bs)
648 struct super_block *sb = inode->i_sb;
651 ea_idebug(inode, "name=%d.%s, value=%p, value_len=%ld",
652 i->name_index, i->name, i->value, (long)i->value_len);
654 if (EXT4_I(inode)->i_file_acl) {
655 /* The inode already has an extended attribute block. */
656 bs->bh = sb_bread(sb, EXT4_I(inode)->i_file_acl);
660 ea_bdebug(bs->bh, "b_count=%d, refcount=%d",
661 atomic_read(&(bs->bh->b_count)),
662 le32_to_cpu(BHDR(bs->bh)->h_refcount));
663 if (ext4_xattr_check_block(bs->bh)) {
664 ext4_error(sb, __func__,
665 "inode %lu: bad block %llu", inode->i_ino,
666 EXT4_I(inode)->i_file_acl);
670 /* Find the named attribute. */
671 bs->s.base = BHDR(bs->bh);
672 bs->s.first = BFIRST(bs->bh);
673 bs->s.end = bs->bh->b_data + bs->bh->b_size;
674 bs->s.here = bs->s.first;
675 error = ext4_xattr_find_entry(&bs->s.here, i->name_index,
676 i->name, bs->bh->b_size, 1);
677 if (error && error != -ENODATA)
679 bs->s.not_found = error;
688 ext4_xattr_block_set(handle_t *handle, struct inode *inode,
689 struct ext4_xattr_info *i,
690 struct ext4_xattr_block_find *bs)
692 unsigned long count = 1;
693 struct super_block *sb = inode->i_sb;
694 struct buffer_head *new_bh = NULL;
695 struct ext4_xattr_search *s = &bs->s;
696 struct mb_cache_entry *ce = NULL;
699 #define header(x) ((struct ext4_xattr_header *)(x))
701 if (i->value && i->value_len > sb->s_blocksize)
704 ce = mb_cache_entry_get(ext4_xattr_cache, bs->bh->b_bdev,
706 error = ext4_journal_get_write_access(handle, bs->bh);
711 if (header(s->base)->h_refcount == cpu_to_le32(1)) {
713 mb_cache_entry_free(ce);
716 ea_bdebug(bs->bh, "modifying in-place");
717 error = ext4_xattr_set_entry(i, s);
719 if (!IS_LAST_ENTRY(s->first))
720 ext4_xattr_rehash(header(s->base),
722 ext4_xattr_cache_insert(bs->bh);
724 unlock_buffer(bs->bh);
728 error = ext4_journal_dirty_metadata(handle,
734 int offset = (char *)s->here - bs->bh->b_data;
736 unlock_buffer(bs->bh);
737 jbd2_journal_release_buffer(handle, bs->bh);
739 mb_cache_entry_release(ce);
742 ea_bdebug(bs->bh, "cloning");
743 s->base = kmalloc(bs->bh->b_size, GFP_NOFS);
747 memcpy(s->base, BHDR(bs->bh), bs->bh->b_size);
748 s->first = ENTRY(header(s->base)+1);
749 header(s->base)->h_refcount = cpu_to_le32(1);
750 s->here = ENTRY(s->base + offset);
751 s->end = s->base + bs->bh->b_size;
754 /* Allocate a buffer where we construct the new block. */
755 s->base = kzalloc(sb->s_blocksize, GFP_NOFS);
756 /* assert(header == s->base) */
760 header(s->base)->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
761 header(s->base)->h_blocks = cpu_to_le32(1);
762 header(s->base)->h_refcount = cpu_to_le32(1);
763 s->first = ENTRY(header(s->base)+1);
764 s->here = ENTRY(header(s->base)+1);
765 s->end = s->base + sb->s_blocksize;
768 error = ext4_xattr_set_entry(i, s);
773 if (!IS_LAST_ENTRY(s->first))
774 ext4_xattr_rehash(header(s->base), s->here);
777 if (!IS_LAST_ENTRY(s->first)) {
778 new_bh = ext4_xattr_cache_find(inode, header(s->base), &ce);
780 /* We found an identical block in the cache. */
781 if (new_bh == bs->bh)
782 ea_bdebug(new_bh, "keeping");
784 /* The old block is released after updating
787 if (DQUOT_ALLOC_BLOCK(inode, 1))
789 error = ext4_journal_get_write_access(handle,
794 le32_add_cpu(&BHDR(new_bh)->h_refcount, 1);
795 ea_bdebug(new_bh, "reusing; refcount now=%d",
796 le32_to_cpu(BHDR(new_bh)->h_refcount));
797 unlock_buffer(new_bh);
798 error = ext4_journal_dirty_metadata(handle,
803 mb_cache_entry_release(ce);
805 } else if (bs->bh && s->base == bs->bh->b_data) {
806 /* We were modifying this block in-place. */
807 ea_bdebug(bs->bh, "keeping this block");
811 /* We need to allocate a new block */
812 ext4_fsblk_t goal = ext4_group_first_block_no(sb,
813 EXT4_I(inode)->i_block_group);
814 ext4_fsblk_t block = ext4_new_meta_blocks(handle, inode,
815 goal, &count, &error);
818 ea_idebug(inode, "creating block %d", block);
820 new_bh = sb_getblk(sb, block);
823 ext4_free_blocks(handle, inode, block, 1, 1);
828 error = ext4_journal_get_create_access(handle, new_bh);
830 unlock_buffer(new_bh);
833 memcpy(new_bh->b_data, s->base, new_bh->b_size);
834 set_buffer_uptodate(new_bh);
835 unlock_buffer(new_bh);
836 ext4_xattr_cache_insert(new_bh);
837 error = ext4_journal_dirty_metadata(handle, new_bh);
843 /* Update the inode. */
844 EXT4_I(inode)->i_file_acl = new_bh ? new_bh->b_blocknr : 0;
846 /* Drop the previous xattr block. */
847 if (bs->bh && bs->bh != new_bh)
848 ext4_xattr_release_block(handle, inode, bs->bh);
853 mb_cache_entry_release(ce);
855 if (!(bs->bh && s->base == bs->bh->b_data))
861 DQUOT_FREE_BLOCK(inode, 1);
865 ext4_error(inode->i_sb, __func__,
866 "inode %lu: bad block %llu", inode->i_ino,
867 EXT4_I(inode)->i_file_acl);
873 struct ext4_xattr_ibody_find {
874 struct ext4_xattr_search s;
875 struct ext4_iloc iloc;
879 ext4_xattr_ibody_find(struct inode *inode, struct ext4_xattr_info *i,
880 struct ext4_xattr_ibody_find *is)
882 struct ext4_xattr_ibody_header *header;
883 struct ext4_inode *raw_inode;
886 if (EXT4_I(inode)->i_extra_isize == 0)
888 raw_inode = ext4_raw_inode(&is->iloc);
889 header = IHDR(inode, raw_inode);
890 is->s.base = is->s.first = IFIRST(header);
891 is->s.here = is->s.first;
892 is->s.end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
893 if (EXT4_I(inode)->i_state & EXT4_STATE_XATTR) {
894 error = ext4_xattr_check_names(IFIRST(header), is->s.end);
897 /* Find the named attribute. */
898 error = ext4_xattr_find_entry(&is->s.here, i->name_index,
900 (void *)is->s.base, 0);
901 if (error && error != -ENODATA)
903 is->s.not_found = error;
909 ext4_xattr_ibody_set(handle_t *handle, struct inode *inode,
910 struct ext4_xattr_info *i,
911 struct ext4_xattr_ibody_find *is)
913 struct ext4_xattr_ibody_header *header;
914 struct ext4_xattr_search *s = &is->s;
917 if (EXT4_I(inode)->i_extra_isize == 0)
919 error = ext4_xattr_set_entry(i, s);
922 header = IHDR(inode, ext4_raw_inode(&is->iloc));
923 if (!IS_LAST_ENTRY(s->first)) {
924 header->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
925 EXT4_I(inode)->i_state |= EXT4_STATE_XATTR;
927 header->h_magic = cpu_to_le32(0);
928 EXT4_I(inode)->i_state &= ~EXT4_STATE_XATTR;
934 * ext4_xattr_set_handle()
936 * Create, replace or remove an extended attribute for this inode. Buffer
937 * is NULL to remove an existing extended attribute, and non-NULL to
938 * either replace an existing extended attribute, or create a new extended
939 * attribute. The flags XATTR_REPLACE and XATTR_CREATE
940 * specify that an extended attribute must exist and must not exist
941 * previous to the call, respectively.
943 * Returns 0, or a negative error number on failure.
946 ext4_xattr_set_handle(handle_t *handle, struct inode *inode, int name_index,
947 const char *name, const void *value, size_t value_len,
950 struct ext4_xattr_info i = {
951 .name_index = name_index,
954 .value_len = value_len,
957 struct ext4_xattr_ibody_find is = {
958 .s = { .not_found = -ENODATA, },
960 struct ext4_xattr_block_find bs = {
961 .s = { .not_found = -ENODATA, },
963 unsigned long no_expand;
968 if (strlen(name) > 255)
970 down_write(&EXT4_I(inode)->xattr_sem);
971 no_expand = EXT4_I(inode)->i_state & EXT4_STATE_NO_EXPAND;
972 EXT4_I(inode)->i_state |= EXT4_STATE_NO_EXPAND;
974 error = ext4_get_inode_loc(inode, &is.iloc);
978 if (EXT4_I(inode)->i_state & EXT4_STATE_NEW) {
979 struct ext4_inode *raw_inode = ext4_raw_inode(&is.iloc);
980 memset(raw_inode, 0, EXT4_SB(inode->i_sb)->s_inode_size);
981 EXT4_I(inode)->i_state &= ~EXT4_STATE_NEW;
984 error = ext4_xattr_ibody_find(inode, &i, &is);
988 error = ext4_xattr_block_find(inode, &i, &bs);
991 if (is.s.not_found && bs.s.not_found) {
993 if (flags & XATTR_REPLACE)
1000 if (flags & XATTR_CREATE)
1003 error = ext4_journal_get_write_access(handle, is.iloc.bh);
1007 if (!is.s.not_found)
1008 error = ext4_xattr_ibody_set(handle, inode, &i, &is);
1009 else if (!bs.s.not_found)
1010 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1012 error = ext4_xattr_ibody_set(handle, inode, &i, &is);
1013 if (!error && !bs.s.not_found) {
1015 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1016 } else if (error == -ENOSPC) {
1017 if (EXT4_I(inode)->i_file_acl && !bs.s.base) {
1018 error = ext4_xattr_block_find(inode, &i, &bs);
1022 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1025 if (!is.s.not_found) {
1027 error = ext4_xattr_ibody_set(handle, inode, &i,
1033 ext4_xattr_update_super_block(handle, inode->i_sb);
1034 inode->i_ctime = ext4_current_time(inode);
1036 EXT4_I(inode)->i_state &= ~EXT4_STATE_NO_EXPAND;
1037 error = ext4_mark_iloc_dirty(handle, inode, &is.iloc);
1039 * The bh is consumed by ext4_mark_iloc_dirty, even with
1051 EXT4_I(inode)->i_state &= ~EXT4_STATE_NO_EXPAND;
1052 up_write(&EXT4_I(inode)->xattr_sem);
1059 * Like ext4_xattr_set_handle, but start from an inode. This extended
1060 * attribute modification is a filesystem transaction by itself.
1062 * Returns 0, or a negative error number on failure.
1065 ext4_xattr_set(struct inode *inode, int name_index, const char *name,
1066 const void *value, size_t value_len, int flags)
1069 int error, retries = 0;
1072 handle = ext4_journal_start(inode, EXT4_DATA_TRANS_BLOCKS(inode->i_sb));
1073 if (IS_ERR(handle)) {
1074 error = PTR_ERR(handle);
1078 error = ext4_xattr_set_handle(handle, inode, name_index, name,
1079 value, value_len, flags);
1080 error2 = ext4_journal_stop(handle);
1081 if (error == -ENOSPC &&
1082 ext4_should_retry_alloc(inode->i_sb, &retries))
1092 * Shift the EA entries in the inode to create space for the increased
1095 static void ext4_xattr_shift_entries(struct ext4_xattr_entry *entry,
1096 int value_offs_shift, void *to,
1097 void *from, size_t n, int blocksize)
1099 struct ext4_xattr_entry *last = entry;
1102 /* Adjust the value offsets of the entries */
1103 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
1104 if (!last->e_value_block && last->e_value_size) {
1105 new_offs = le16_to_cpu(last->e_value_offs) +
1107 BUG_ON(new_offs + le32_to_cpu(last->e_value_size)
1109 last->e_value_offs = cpu_to_le16(new_offs);
1112 /* Shift the entries by n bytes */
1113 memmove(to, from, n);
1117 * Expand an inode by new_extra_isize bytes when EAs are present.
1118 * Returns 0 on success or negative error number on failure.
1120 int ext4_expand_extra_isize_ea(struct inode *inode, int new_extra_isize,
1121 struct ext4_inode *raw_inode, handle_t *handle)
1123 struct ext4_xattr_ibody_header *header;
1124 struct ext4_xattr_entry *entry, *last, *first;
1125 struct buffer_head *bh = NULL;
1126 struct ext4_xattr_ibody_find *is = NULL;
1127 struct ext4_xattr_block_find *bs = NULL;
1128 char *buffer = NULL, *b_entry_name = NULL;
1129 size_t min_offs, free;
1130 int total_ino, total_blk;
1131 void *base, *start, *end;
1132 int extra_isize = 0, error = 0, tried_min_extra_isize = 0;
1133 int s_min_extra_isize = le16_to_cpu(EXT4_SB(inode->i_sb)->s_es->s_min_extra_isize);
1135 down_write(&EXT4_I(inode)->xattr_sem);
1137 if (EXT4_I(inode)->i_extra_isize >= new_extra_isize) {
1138 up_write(&EXT4_I(inode)->xattr_sem);
1142 header = IHDR(inode, raw_inode);
1143 entry = IFIRST(header);
1146 * Check if enough free space is available in the inode to shift the
1147 * entries ahead by new_extra_isize.
1150 base = start = entry;
1151 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
1152 min_offs = end - base;
1154 total_ino = sizeof(struct ext4_xattr_ibody_header);
1156 free = ext4_xattr_free_space(last, &min_offs, base, &total_ino);
1157 if (free >= new_extra_isize) {
1158 entry = IFIRST(header);
1159 ext4_xattr_shift_entries(entry, EXT4_I(inode)->i_extra_isize
1160 - new_extra_isize, (void *)raw_inode +
1161 EXT4_GOOD_OLD_INODE_SIZE + new_extra_isize,
1162 (void *)header, total_ino,
1163 inode->i_sb->s_blocksize);
1164 EXT4_I(inode)->i_extra_isize = new_extra_isize;
1170 * Enough free space isn't available in the inode, check if
1171 * EA block can hold new_extra_isize bytes.
1173 if (EXT4_I(inode)->i_file_acl) {
1174 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
1178 if (ext4_xattr_check_block(bh)) {
1179 ext4_error(inode->i_sb, __func__,
1180 "inode %lu: bad block %llu", inode->i_ino,
1181 EXT4_I(inode)->i_file_acl);
1187 end = bh->b_data + bh->b_size;
1188 min_offs = end - base;
1189 free = ext4_xattr_free_space(first, &min_offs, base,
1191 if (free < new_extra_isize) {
1192 if (!tried_min_extra_isize && s_min_extra_isize) {
1193 tried_min_extra_isize++;
1194 new_extra_isize = s_min_extra_isize;
1202 free = inode->i_sb->s_blocksize;
1205 while (new_extra_isize > 0) {
1206 size_t offs, size, entry_size;
1207 struct ext4_xattr_entry *small_entry = NULL;
1208 struct ext4_xattr_info i = {
1212 unsigned int total_size; /* EA entry size + value size */
1213 unsigned int shift_bytes; /* No. of bytes to shift EAs by? */
1214 unsigned int min_total_size = ~0U;
1216 is = kzalloc(sizeof(struct ext4_xattr_ibody_find), GFP_NOFS);
1217 bs = kzalloc(sizeof(struct ext4_xattr_block_find), GFP_NOFS);
1223 is->s.not_found = -ENODATA;
1224 bs->s.not_found = -ENODATA;
1228 last = IFIRST(header);
1229 /* Find the entry best suited to be pushed into EA block */
1231 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
1233 EXT4_XATTR_SIZE(le32_to_cpu(last->e_value_size)) +
1234 EXT4_XATTR_LEN(last->e_name_len);
1235 if (total_size <= free && total_size < min_total_size) {
1236 if (total_size < new_extra_isize) {
1240 min_total_size = total_size;
1245 if (entry == NULL) {
1247 entry = small_entry;
1249 if (!tried_min_extra_isize &&
1250 s_min_extra_isize) {
1251 tried_min_extra_isize++;
1252 new_extra_isize = s_min_extra_isize;
1259 offs = le16_to_cpu(entry->e_value_offs);
1260 size = le32_to_cpu(entry->e_value_size);
1261 entry_size = EXT4_XATTR_LEN(entry->e_name_len);
1262 i.name_index = entry->e_name_index,
1263 buffer = kmalloc(EXT4_XATTR_SIZE(size), GFP_NOFS);
1264 b_entry_name = kmalloc(entry->e_name_len + 1, GFP_NOFS);
1265 if (!buffer || !b_entry_name) {
1269 /* Save the entry name and the entry value */
1270 memcpy(buffer, (void *)IFIRST(header) + offs,
1271 EXT4_XATTR_SIZE(size));
1272 memcpy(b_entry_name, entry->e_name, entry->e_name_len);
1273 b_entry_name[entry->e_name_len] = '\0';
1274 i.name = b_entry_name;
1276 error = ext4_get_inode_loc(inode, &is->iloc);
1280 error = ext4_xattr_ibody_find(inode, &i, is);
1284 /* Remove the chosen entry from the inode */
1285 error = ext4_xattr_ibody_set(handle, inode, &i, is);
1287 entry = IFIRST(header);
1288 if (entry_size + EXT4_XATTR_SIZE(size) >= new_extra_isize)
1289 shift_bytes = new_extra_isize;
1291 shift_bytes = entry_size + size;
1292 /* Adjust the offsets and shift the remaining entries ahead */
1293 ext4_xattr_shift_entries(entry, EXT4_I(inode)->i_extra_isize -
1294 shift_bytes, (void *)raw_inode +
1295 EXT4_GOOD_OLD_INODE_SIZE + extra_isize + shift_bytes,
1296 (void *)header, total_ino - entry_size,
1297 inode->i_sb->s_blocksize);
1299 extra_isize += shift_bytes;
1300 new_extra_isize -= shift_bytes;
1301 EXT4_I(inode)->i_extra_isize = extra_isize;
1303 i.name = b_entry_name;
1306 error = ext4_xattr_block_find(inode, &i, bs);
1310 /* Add entry which was removed from the inode into the block */
1311 error = ext4_xattr_block_set(handle, inode, &i, bs);
1314 kfree(b_entry_name);
1316 brelse(is->iloc.bh);
1321 up_write(&EXT4_I(inode)->xattr_sem);
1325 kfree(b_entry_name);
1328 brelse(is->iloc.bh);
1332 up_write(&EXT4_I(inode)->xattr_sem);
1339 * ext4_xattr_delete_inode()
1341 * Free extended attribute resources associated with this inode. This
1342 * is called immediately before an inode is freed. We have exclusive
1343 * access to the inode.
1346 ext4_xattr_delete_inode(handle_t *handle, struct inode *inode)
1348 struct buffer_head *bh = NULL;
1350 if (!EXT4_I(inode)->i_file_acl)
1352 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
1354 ext4_error(inode->i_sb, __func__,
1355 "inode %lu: block %llu read error", inode->i_ino,
1356 EXT4_I(inode)->i_file_acl);
1359 if (BHDR(bh)->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC) ||
1360 BHDR(bh)->h_blocks != cpu_to_le32(1)) {
1361 ext4_error(inode->i_sb, __func__,
1362 "inode %lu: bad block %llu", inode->i_ino,
1363 EXT4_I(inode)->i_file_acl);
1366 ext4_xattr_release_block(handle, inode, bh);
1367 EXT4_I(inode)->i_file_acl = 0;
1374 * ext4_xattr_put_super()
1376 * This is called when a file system is unmounted.
1379 ext4_xattr_put_super(struct super_block *sb)
1381 mb_cache_shrink(sb->s_bdev);
1385 * ext4_xattr_cache_insert()
1387 * Create a new entry in the extended attribute cache, and insert
1388 * it unless such an entry is already in the cache.
1390 * Returns 0, or a negative error number on failure.
1393 ext4_xattr_cache_insert(struct buffer_head *bh)
1395 __u32 hash = le32_to_cpu(BHDR(bh)->h_hash);
1396 struct mb_cache_entry *ce;
1399 ce = mb_cache_entry_alloc(ext4_xattr_cache, GFP_NOFS);
1401 ea_bdebug(bh, "out of memory");
1404 error = mb_cache_entry_insert(ce, bh->b_bdev, bh->b_blocknr, &hash);
1406 mb_cache_entry_free(ce);
1407 if (error == -EBUSY) {
1408 ea_bdebug(bh, "already in cache");
1412 ea_bdebug(bh, "inserting [%x]", (int)hash);
1413 mb_cache_entry_release(ce);
1420 * Compare two extended attribute blocks for equality.
1422 * Returns 0 if the blocks are equal, 1 if they differ, and
1423 * a negative error number on errors.
1426 ext4_xattr_cmp(struct ext4_xattr_header *header1,
1427 struct ext4_xattr_header *header2)
1429 struct ext4_xattr_entry *entry1, *entry2;
1431 entry1 = ENTRY(header1+1);
1432 entry2 = ENTRY(header2+1);
1433 while (!IS_LAST_ENTRY(entry1)) {
1434 if (IS_LAST_ENTRY(entry2))
1436 if (entry1->e_hash != entry2->e_hash ||
1437 entry1->e_name_index != entry2->e_name_index ||
1438 entry1->e_name_len != entry2->e_name_len ||
1439 entry1->e_value_size != entry2->e_value_size ||
1440 memcmp(entry1->e_name, entry2->e_name, entry1->e_name_len))
1442 if (entry1->e_value_block != 0 || entry2->e_value_block != 0)
1444 if (memcmp((char *)header1 + le16_to_cpu(entry1->e_value_offs),
1445 (char *)header2 + le16_to_cpu(entry2->e_value_offs),
1446 le32_to_cpu(entry1->e_value_size)))
1449 entry1 = EXT4_XATTR_NEXT(entry1);
1450 entry2 = EXT4_XATTR_NEXT(entry2);
1452 if (!IS_LAST_ENTRY(entry2))
1458 * ext4_xattr_cache_find()
1460 * Find an identical extended attribute block.
1462 * Returns a pointer to the block found, or NULL if such a block was
1463 * not found or an error occurred.
1465 static struct buffer_head *
1466 ext4_xattr_cache_find(struct inode *inode, struct ext4_xattr_header *header,
1467 struct mb_cache_entry **pce)
1469 __u32 hash = le32_to_cpu(header->h_hash);
1470 struct mb_cache_entry *ce;
1472 if (!header->h_hash)
1473 return NULL; /* never share */
1474 ea_idebug(inode, "looking for cached blocks [%x]", (int)hash);
1476 ce = mb_cache_entry_find_first(ext4_xattr_cache, 0,
1477 inode->i_sb->s_bdev, hash);
1479 struct buffer_head *bh;
1482 if (PTR_ERR(ce) == -EAGAIN)
1486 bh = sb_bread(inode->i_sb, ce->e_block);
1488 ext4_error(inode->i_sb, __func__,
1489 "inode %lu: block %lu read error",
1490 inode->i_ino, (unsigned long) ce->e_block);
1491 } else if (le32_to_cpu(BHDR(bh)->h_refcount) >=
1492 EXT4_XATTR_REFCOUNT_MAX) {
1493 ea_idebug(inode, "block %lu refcount %d>=%d",
1494 (unsigned long) ce->e_block,
1495 le32_to_cpu(BHDR(bh)->h_refcount),
1496 EXT4_XATTR_REFCOUNT_MAX);
1497 } else if (ext4_xattr_cmp(header, BHDR(bh)) == 0) {
1502 ce = mb_cache_entry_find_next(ce, 0, inode->i_sb->s_bdev, hash);
1507 #define NAME_HASH_SHIFT 5
1508 #define VALUE_HASH_SHIFT 16
1511 * ext4_xattr_hash_entry()
1513 * Compute the hash of an extended attribute.
1515 static inline void ext4_xattr_hash_entry(struct ext4_xattr_header *header,
1516 struct ext4_xattr_entry *entry)
1519 char *name = entry->e_name;
1522 for (n = 0; n < entry->e_name_len; n++) {
1523 hash = (hash << NAME_HASH_SHIFT) ^
1524 (hash >> (8*sizeof(hash) - NAME_HASH_SHIFT)) ^
1528 if (entry->e_value_block == 0 && entry->e_value_size != 0) {
1529 __le32 *value = (__le32 *)((char *)header +
1530 le16_to_cpu(entry->e_value_offs));
1531 for (n = (le32_to_cpu(entry->e_value_size) +
1532 EXT4_XATTR_ROUND) >> EXT4_XATTR_PAD_BITS; n; n--) {
1533 hash = (hash << VALUE_HASH_SHIFT) ^
1534 (hash >> (8*sizeof(hash) - VALUE_HASH_SHIFT)) ^
1535 le32_to_cpu(*value++);
1538 entry->e_hash = cpu_to_le32(hash);
1541 #undef NAME_HASH_SHIFT
1542 #undef VALUE_HASH_SHIFT
1544 #define BLOCK_HASH_SHIFT 16
1547 * ext4_xattr_rehash()
1549 * Re-compute the extended attribute hash value after an entry has changed.
1551 static void ext4_xattr_rehash(struct ext4_xattr_header *header,
1552 struct ext4_xattr_entry *entry)
1554 struct ext4_xattr_entry *here;
1557 ext4_xattr_hash_entry(header, entry);
1558 here = ENTRY(header+1);
1559 while (!IS_LAST_ENTRY(here)) {
1560 if (!here->e_hash) {
1561 /* Block is not shared if an entry's hash value == 0 */
1565 hash = (hash << BLOCK_HASH_SHIFT) ^
1566 (hash >> (8*sizeof(hash) - BLOCK_HASH_SHIFT)) ^
1567 le32_to_cpu(here->e_hash);
1568 here = EXT4_XATTR_NEXT(here);
1570 header->h_hash = cpu_to_le32(hash);
1573 #undef BLOCK_HASH_SHIFT
1576 init_ext4_xattr(void)
1578 ext4_xattr_cache = mb_cache_create("ext4_xattr", NULL,
1579 sizeof(struct mb_cache_entry) +
1580 sizeof(((struct mb_cache_entry *) 0)->e_indexes[0]), 1, 6);
1581 if (!ext4_xattr_cache)
1587 exit_ext4_xattr(void)
1589 if (ext4_xattr_cache)
1590 mb_cache_destroy(ext4_xattr_cache);
1591 ext4_xattr_cache = NULL;